DETERMINANTS INFLUENCING THE ADOPTION OF ARTIFICIAL INTELLIGENCE TECHNOLOGY IN NON-LIFE INSURERS

How to cite this paper: Nguyen, T. H. D., Nguyen, X. T., Le, T. H. T., & Bui, Q. A. (2024). Determinants influencing the adoption of artificial intelligence technology in non-life insurers. Corporate Governance and Organizational Behavior Review, 8 (1), 205–212. https


INTRODUCTION
Artificial intelligence (AI), originally developed in the 1950s, is known as a computer system whose intelligence is equal to that of a human (Eling et al., 2022).Since AI has the power to enhance and improve organizational performance by solving complex business challenges, it is adopted in various industries, such as manufacturing, education, food service, construction, tourism, transportation, finance and insurance, etc.More and more enterprises are using AI applications in all operating activities, making AI a leading strategic technology (Phuoc, 2022).
Recently, adopting AI technology has become a popular trend in the insurance industry all over the world.According to Ceylan (2022), the vast majority of insurers currently have several activities in relation to AI technology, with 30% actively looking at opportunities, and another 50% fully consistently investing in new AI technology.Thanks to the ability to perform complex computational tasks, AI technology is rapidly transforming various financial services industries, particularly within nonlife insurance.AI has a broad effect along the entire value chain of insurance from product design to underwriting and actuarial activities and claims management (Xu & Zweifel, 2020).Adopting AI technology leads to both pros and cons for insurance firms.On the one hand, AI-enabled applications could be used to deal with numerous policy enrollments and claim settlements, build new products, enhance customer experience, etc.On the other hand, several potential risks also arise, such as cyber-attacks and identity theft.
Although factors affecting AI adoption in many countries are intensively investigated ( Na et al., 2023), the studies for Vietnam are still sparse.This paper contributes to the literature review on AI adoption in Vietnam, the second fastest-growing economy in Southeast Asia.After two decades of development, the Vietnamese non-life insurance market thrived dramatically with a growth rate in revenue of nearly 15% per year.At the end of 2022, the revenue of non-life insurance reached approximately $3.7 billion, accounting for 1% of the national gross domestic product (GDP).Due to the current mainstream of digital transformation in Vietnam, both state agencies and insurance firms expect that AI technology will boom in the following years.However, according to Pham et al. (2022), less than 10% of Vietnamese companies are actively developing their AI applications.Many insurance employees are unwilling to use AI technology due to potential risks, ease of use, facilitating conditions, etc.Hence, we aim to provide an adequate understanding of AI adoption in non-life insurers and its key drivers in Vietnam.This research conducts a survey among employees in Vietnamese non-life insurance firms, who directly use AI technology for their work.
Based on data from nearly 470 questionnaires, we find that the behavioral intention to adopt AI technology is negatively impacted by perceived risk.Its coefficient is only approximately -0.2, indicating that insurance employees do not pay a large amount of attention to perceived risk.The strongest determinant of AI adoption in Vietnamese non-life is attitudes toward adoption, which is positively related to perceived usefulness and perceived ease of use.Finally, facilitating conditions and perceived selfefficacy of employees also play essential roles in adopting AI technology in Vietnamese non-life insurers.
The rest of this article is structured as follows.Section 2 reviews the relevant literature.Section 3 describes the data and methodology.Section 4 presents and discusses the empirical results, then Section 5 draws the conclusions.2023) document that satisfaction with AI technology is linked to the perceived ease of use and perceived usefulness, leading to higher intention to adopt AI technology in Korean construction firms.In the insurance industry, Gupta et al. (2022) state that managerial capability and usefulness significantly predict behavioral intention of adopting AI in India.

LITERATURE REVIEW AND HYPOTHESES DEVELOPMENT
The impact of AI on the insurance industry in developed markets is also comprehensively investigated.According to Eling et al. (2022), AI adoption not only creates new revenue streams for insurers but also enhances cost efficiency.Computer systems with AI are able to recognize patterns in the insured's data, and then procedure a more accurate estimation of loss likelihood.As a result, insurance firms can automate the underwriting processes and manage claim settlements more effectively.Furthermore, thanks to individual information collected by AI, insurance products are tailored to each customer and their current situation, enhancing their experience (Zarifis & Cheng, 2022).The application of AI in insurance also creates several potential risks.To recognize patterns, computer systems should be trained with a large amount of data, which could be a vulnerable target for cyber-attacks.More dangerously, the insured's data includes individual information such as bank accounts, national identity, social security numbers, etc.Furthermore, insufficient data leads to biased estimation of loss probabilities.Then, the underwriting processes might be erroneous.
To the best of our knowledge, in-depth research on the adoption of AI technology in Vietnam is still limited.Phuoc (2022) examines AI adoption by surveying nearly 200 managers in Vietnamese firms.Empirical results show that the technical complexity (or perceived ease of use), top management support, and usefulness are critical factors affecting the adoption of AI in Vietnam.

Pham et al. (2022) investigate the intention to use insurance applications of Vietnamese customers.
Based on 166 questionnaires, they find that usefulness, ease of use, and trust have a positive impact on the intention to use insurance applications.Meanwhile, perceived risk seems to be an insignificant factor.Since there is no published paper analyzing the adoption of AI technology in the Vietnamese insurance industry, we contribute to the literature review on AI adoption in Vietnam.Hypotheses are built as follows.
Perceived risk (PR): Various papers indicate that users' perceived risk has an important impact on the adoption of AI technology (Cukurova et al., 2020;Horowitz & Kahn, 2021).According to Horowitz and Kahn (2021), perceived risk negatively influences AI adoption.
H1: Perceived risk negatively influences the behavioral intention to adopt artificial intelligence technology in Vietnamese non-life insurance companies.
Perceived self-efficacy (PSE): According to Kulviwat et al. ( 2014), self-efficacy plays an essential role in shaping the impact of cognition and its effects on high technology adoption.Li et al. (2022) prove that perceived self-efficacy positively supports the perceived usefulness of AI technology applications.
H2: Perceived self-efficacy of employees in insurance firms positively influences the perceived usefulness of artificial intelligence technology in Vietnamese non-life insurance companies.
Facilitating condition (FC): Alam et al. (2020) define facilitating condition as the availability of necessary infrastructure and organization to adopt the AI-powered system.For example, user-friendly computing infrastructure and guidelines to use AIbased software fuel AI adoption.Grover et al. ( 2022) also imply that the facilitating condition is the key determinant affecting the perceived ease of use of technology users.
H3: Facilitating conditions positively influence the perceived ease of use of artificial intelligence technology in Vietnamese non-life insurance companies.
Perceived ease of use (PEOU): Perceived ease of use measures how well the user adopts an AI-based software without effort, known as one of the major variables of the technology acceptance model.Perceived ease of use is confirmed to influence the perceived usefulness as well as the behavioral intention to adopt AI technology in the existing literature (Chen et  H4: Perceived ease of use positively influences the perceived usefulness of artificial intelligence technology in Vietnamese non-life insurance companies.
H5: Perceived ease of use positively influences attitudes toward adopting artificial intelligence technology in Vietnamese non-life insurance companies.
Perceived usefulness (PU): Perceived usefulness is the degree to which users believe that adopting AI technology would enhance their productivity.Perceived usefulness has a positive effect on the attitude to adopt AI technology, as reported in several studies (Li et  H7: Attitudes toward adoption positively influence the behavioral intention to adopt artificial intelligence technology in Vietnamese non-life insurance companies.

RESEARCH METHODOLOGY
This study aims to examine determinants influencing the adoption of AI technology in Vietnamese non-life insurance companies.Based on the literature review, we propose a framework of factors affecting AI technology adoption.A pilot study with 15 experts is initially used to redefine determinants as well as test the validity of the questionnaire.The final questionnaire includes two parts.The first part provides demographic information of respondents.The second part consists of questions related to factors influencing the adoption of AI technology.Then, we designed a questionnaire in Google Forms and sent it to nearly 500 employees in Vietnamese non-life insurance companies from April 5 to April 25, 2023; 468 fulfilled responses meeting the preset requirements are collected.Demographics of respondents are given in Table 1.In terms of gender, the majority of respondents are male (54.5%) and their ages range between 40 and 49 years old.The highest educational attainment belongs to college/university, at nearly 75%.In terms of positions, about 84% of respondents are employees.The research process is summarized in Figure 2. The first step is descriptive analysis, which describes samples in terms of variables.Secondly, we test the reliability of measurement scales with Cronbach's alpha.The two next steps are exploratory factor analysis (EFA) and confirmatory factor analysis (CFA).Finally, we test hypotheses by running the structural equation model (SEM).The CMIN/DF stands for Chi-square statistics/degree of freedom, implying discrepancy divided by degree of freedom.The GFI stands for goodness of fit index.The AGFI stands for adjusted goodness of fit index.The CFI stands for comparative fit index.The RMSEA stands for root mean square error of approximation.Acceptable ranges are collected from Hair et al. (2019).

Structural model and hypotheses testing
According to Table 6, all the fit values for the structural model are ranged in the recommended values.
Therefore, the structural model achieves a good level of fit.The SEM is performed to test the hypotheses proposed in Section 2. Results are summarized in Figure 3 and Table 7.

Discussion
In this paper, seven hypotheses are tested and all of them received support from the data sample.
The first hypothesis H1 is confirmed.The perceived risk (β = -0.226,p-value = 0.000) has a significantly negative impact on the behavioural intention to adopt AI technology in non-life insurers.To adopt AI technology, insurers must collect, handle, and store a huge amount of sensitive customer data, leading to several potential risks such as cyber-attacks and identity theft.Therefore, if the perceived risk increases by 1 unit, the behavioural intention to adopt AI technology in non-life insurers would decline by 0.226.
Secondly, the perceived self-efficacy of employees (β = 0.265, p-value = 0.000) has a positive influence on the perceived usefulness of AI technology.Additionally, if insurance companies create favorable conditions for their employees to improve their knowledge of AI technology, they would also easily use AI technology.Therefore, the third hypothesis H3 is confirmed.Facilitating conditions positively affect the perceived ease of use of AI technology (β = 0.245, p-value = 0.000).If nonlife insurers invest in technical and organizational infrastructures, the perceived ease of use of AI technology would be enhanced.This result is in line with Alam et al. (2020) and Islam et al. (2022).
According to Table 7, perceived ease of use positively influences the perceived usefulness of AI technology (β = 0.284, p-value = 0.000).User-friendly AI technology not only makes employees satisfied but also improves their performance, thus increasing their perceived usefulness.
As expected, perceived ease of use is found to be positively related to attitudes toward adoption, with a significant coefficient of nearly 0.3.If employees have excellent knowledge and skills in AI technology, they believe in their own ability to exploit the benefits of AI technology to enhance their working performance.Consequently, they tend to use AI technology frequently, thereby increasing their awareness of the usefulness of AI technology.In contrast, if employees do not know about AI technology, they will find it difficult to use AI technology.As a result, they will tend to use this technology infrequently, reducing their awareness of the usefulness of AI technology.It is consistent with the results of Hong (2022).
With regard to the effect of perceived usefulness, H6 pertaining to the positive relationship between PU and AT toward behavioural intention to adopt AI technology is supported (β = 0.295, p-value = 0.000).In fact, if employees believe that AI technology is easy to use, their attitude towards adopting AI technology would be positive.Then, all staff members acquiesce to the application of technology in their daily tasks.Meanwhile, employees who believe that adopting AI technology is difficult have feelings of anxiety, leading to a negative attitude towards adopting AI technology.They might suppose that adopting AI technology is inefficient.Finally, the hypothesized path of attitudes toward adoption and the behavioural intention to adopt AI technology is statistically significant (β = 0.484, p-value = 000).The strongest positive effect belongs to the attitudes toward adoption.When insurance companies plan to invest in AI technology, they would take the vote of employees.If most staff members believe that adopting AI technology is beneficial for their company, the behavioural intention to adopt AI technology would increase.On the contrary, if the majority of employees and managers argue that AI technology does not significantly improve their working efficiency, decreasing the behavioural intention to adopt AI technology in non-life insurers.

CONCLUSION
The adoption of AI technology brings various benefits to non-life insurance companies.This research proposes a structural model of factors influencing the adoption of AI technology in Vietnamese non-life insurers.Empirical results show that the behavioural intention to adopt AI technology is negatively impacted by perceived risk, consistent with Cukurova et al. (2020) and Horowitz and Kahn (2021).However, its is less than the coefficient of attitudes toward adoption, implying that perceived risk is not a serious issue for insurance companies.Attitudes toward adoption have the strongest effect on the behavioural intention to adopt AI technology, as suggested by Pham et al. (2022) and Na et al. (2023).Moreover, the paths from perceived usefulness and perceived ease of use to attitudes toward adoption are positively significant.Major insurance employees believe that adopting AI technology would improve their productivity and reduce operating expenses for insurers.Adopting AI technology also enhances the profits of insurance companies if their staff easily apply AI technology to their work.A positive relationship between perceived self-efficacy and perceived usefulness is confirmed, indicating that employees recognize the usefulness of AI technology.We also find a positive relationship between facilitating conditions and perceived ease of use.
Hence, Vietnamese non-life insurers should create more and more facilitating conditions to encourage their employees to adopt AI technology.Some implications could be drawn from the research results.Firstly, since perceived risk has a negative impact on the behavioural intention to adopt AI technology, Vietnamese non-life insurers should pay attention to potential risks in AI adoption.Sensitive client data must be acquired, processed, and stored in a way that complies with security laws, data privacy, and moral and ethical principles.Furthermore, the Department of the Insurance Supervisory Authority should establish legal frameworks for data management in insurance companies.Secondly, facilitating conditions and perceived ease of use are considered preconditions for a successful implementation of AI.Non-life insurers should invest a large amount of funds in building computer systems as well as modern software and applications.These AI applications should be tailored to meet employee preferences, increasing ease of use.Finally, a supportive learning environment where employees are encouraged to explore and experiment with AI applications should be built.Thanks to this environment, employees in insurance firms could collaborate productively with AI applications in their daily tasks such as customer interaction, underwriting, claim settlement, product offerings, etc.
There are several limitations in this paper.Firstly, due to limited time and means of communication, the data sample includes only 468 respondents.Secondly, the paper only examines factors effecting influencing the adoption of AI technology in Vietnamese non-life insurers.The life insurance firms are excluded.Finally, AI technology is relatively unfamiliar and has not been investigated much in Vietnam, leading to several confusions about questions of measurement in questionnaires.
As a result, some directions for further search are suggested.Firstly, expand the data sample to life insurers as well as the number of respondents.Secondly, in-depth interviews with professionals in the field of AI technology should be conducted to make questionnaires more transparent.
With the wide application of AI technology in various industries, there are several papers analyzing factors affecting the adoption of AI.Examining questionnaires of 690 United States officials, Horowitz and Kahn (2021) point out that the behavioral intention to adopt AI technology is negatively affected by perceived risk, especially the theft of personal information through cyber-attacks.Adoption of AI in an organization depends on the computing infrastructure and training of employees, known as facilitating conditions (Grover et al., 2022).According to Islam et al. (2022), the availability of technical and organizational infrastructures, considered facilitating conditions, significantly promotes the behavioral intention to adopt AI technology.Furthermore, the behavioral intention to adopt AI is regarded as a sustainable predictor of actual adoption.Li et al. (2022) state that the behavioral intention to use AI of medical students is predicted by their perceived usefulness and perceived self-efficacy.According to Chen et al. (2022), perceived ease of use and perceived usefulness are the two most significant determinants influencing AI adoption in business-to-business marketing.While these two factors also significantly affect the AI adoption of accounting students in Indonesia (Sudaryanto et al., 2023), technology readiness and digital competence play no role.Na et al. (

Figure 2 .
Figure 2. The research process

Table 1 .
Demographics of respondentsFigure 1.The research framework model

Table 5 .
Rotated component matrix

Table 6 .
Goodness of fit (CFA and SEM)

Table 7 .
Hypotheses testing results